广义可加模型在出生缺陷影响因素分析中的应用及R语言实现过程

Application of generalized additive model in analysis of influencing factors of birth defects and R implementation

目的:探讨广义可加模型在出生缺陷影响因素分析中的应用及R语言实现过程,为实际研究提供方法学探索。方法:对2013年江苏省某3个区县围产儿监测数据拟合广义可加模型,采用Ri386 3.0.3软件mgcv程序包拟合模型,模型拟合包括单因素分析、多因素分析、共曲线性诊断、可视化处理、交互作用分析等。结果:多因素分析表明,体重与出生缺陷的关系类似二次曲线,体重过低过高、早产、男性围产儿发生出生缺陷的危险性都相对较高;共曲线性检验表明体重与身长可能具有共曲线性;交互作用分析表明体重和孕周对出生缺陷的影响可能存在交互作用。结论:R软件mgcv程序包拟合广义可加模型为出生缺陷影响因素的分析提供了灵活有效的工具,有利于实际中的推广应用。

Objective： To explore the application of generalized additive model in analysis of influencing factors of birth defects and R implementation, provide a methodological exploration for actual researches. Methods： mgcv package in Ri386 3.0. 3 software was used for generalized additive model in perinatal monitoring data of three （districts） counties in Jiangsu province in 2013, including univariate analysis, multivariate factor analysis, concurvity diagnosis, visual processing and interaction analysis. Results： Multivariate factor analysis showed that the relationship between weight and birth defect was a quadratic curve, the risks of perinatal infants with high or low birth weight, premature birth and male gender were relatively high; concurvity test showed that weight and height might have a common curve ; interaction analysis indicated that weight and gestational weeks might have interaction effects on birth defects. Conclusion： mgcv package in R software provides flexible and effective tool for generalized additive model in analysis of influencing factors of birth defects, which is helpful to practical promotion and application.